Then

f(x) =
∞
_
−∞
cneinx =
1
2
+
1
iπ
_eix
1
+ e3ix
3
+ e5ix
5
(7.8) + · · ·_
+
1
iπ
_e−ix
−1
+ e−3ix
−3
+ e−5ix
−5
+ · · ·_
.
It is interesting to verify that this is the same as the sine-cosine series we had before.
We could use Euler’s formula for each exponential, but it is easier to collect terms
like this:
f(x) =
1
2
+
2
π
_eix − e−ix
2i
+
1
3
e3ix − e−3ix
2i
(7.9) + · · ·_
=
1
2
+
2
π
_sin x +
1
3
sin 3x + · · ·_
which is the same as (5.12).
PROBLEMS, SECTION 7
1 to 11. Expand the same functions as in Problems 5.1 to 5.11 in Fourier series of
complex exponentials einx on the interval (−π, π) and verify in each case that the answer
is equivalent to the one found in Section 5.
12. PShow that if a real f(x) is expanded in a complex exponential Fourier series ∞
−∞ cneinx, then c−n = ¯cn, where ¯cn means the complex conjugate of cn.
13. If f(x) = 1
2a0 +
P∞
1 an cos nx +
P∞
1 bn sin nx =
P∞
−∞ cneinx, use Euler’s formula
to find an and bn in terms of cn and c−n, and to find cn and c−n in terms of an
and bn.
8. OTHER INTERVALS
The functions sin nx and cos nx and einx have period 2π. We have been considering
(−π, π) as the basic interval of length 2π. Given f(x) on (−π, π), we have first
sketched it for this interval, and then repeated our sketch for the intervals (π, 3π),
(3π, 5π), (−3π,−π), etc. There are (infinitely) many other intervals of length 2π,
any one of which could serve as the basic interval. If we are given f(x) on any
interval of length 2π, we can sketch f(x) for that given basic interval and then
repeat it periodically with period 2π. We thenwant to expand the periodic function
so obtained, in a Fourier series. Recall that in evaluating the Fourier coefficients,
we used average values over a period. The formulas for the coefficients are then
unchanged (except for the limits of integration) if we use other basic intervals of
length 2π. In practice, the intervals (−π, π) and (0, 2π) are the onesmost frequently
used. For f(x) defined on (0, 2π) and then repeated periodically, (5.9), (5.10), and
(7.6) would read
an =
1
π
_ 2π
0
f(x) cos nx dx, bn =
1
π
_ 2π
0
f(x) sinnx dx,
cn =
1
2π
_ 2π
0
f(x)e
−inx dx,and
(5.1) and (7.2) are unchanged.
Figure 8.1
Notice how important it is to sketch a graph to see clearly what function you
are talking about. For example, given f(x) = x2 on (−π, π), the extended function
of period 2π is shown in Figure 8.1. But given f(x) = x2 on (0, 2π), the extended
periodic function is different (see Figure 8.2). On the other hand, given f(x) as in
our example (5.11), or given f(x) = 1 on (0, π), f(x) = 0 on (π, 2π), you can easily
verify by sketching that the graphs of the extended functions are identical. In this
case you would get the same answer from either formulas (5.9), (5.10), and (7.6) or
formulas (8.1).
Figure 8.2
Physics problems do not always come to us with intervals of length 2π. Fortunately,
it is easy now to change to other intervals. Consider intervals of length 2l,
say (−l, l) or (0, 2l). The function sin(nπx/l) has period 2l, since
sin nπ
l
(x + 2l) = sin_nπx
l
+ 2nπ_ = sin nπx
l
.
Similarly, cos(nπx/l) and einπx/l have period 2l. Equations (5.1) and (7.2) are now
replaced by
f(x) = a0
2
+ a1 cos πx
l
+ a2 cos
2πx
l
+・・・
+ b1 sin πx
l
+ b2 sin
2πx
l
+・・・
= a0
2
+
∞
_
1
_an cos nπx
l
+ bn sin nπx
l
_,
f(x) =
∞
_
−∞
cneinπx/l.
(8.2)
362 Fourier Series and Transforms Chapter 7
We have already found the average values over a period of all the functions we
need to use to find an, bn, and cn here. The period is now of length 2l, say −l to l,
so in finding average values of the terms we replace
1
2π
_π
−π
by
1
2l
_l
−l
.
Recall that the average of the square of either the sine or the cosine over a period
is 1
2 and the average of einπx/l ・ e−inπx/l = 1 is 1. Then the formulas (5.9), (5.10),
and (7.6) for the coefficients become
(8.3)
an =
1
l
_l
−l
f(x) cos nπx
l
dx,
bn =
1
l
_l
−l
f(x) sin nπx
l
dx,
cn =
1
2l
_l
−l
f(x)e
−inπx/l dx.
For the basic interval (0, 2l) we need only change the integration limits to 0 to 2l.
The Dirichlet theorem just needs π replaced by l in order to apply here.
Example. Given f(x) =
0, 0 < x < l,
1, l<x<2l.
Expand f(x) in an exponential Fourier series of period 2l. [The function is given
by the same formulas as (5.11) but on a different interval.]
Figure 8.3
First we sketch a graph of f(x) repeated with period 2l (Figure 8.3). By equations
(8.3), we find
cn =
1
2l
_l
0
0 ・ dx +
1
2l
_ 2l
l
1・e
−inπx/l dx
=
1
2l
e−inπx/l
−inπ/l

2l
l
=
1
−2inπ
(e
−2inπ − e
−inπ)
=
1
−2inπ
(1 − einπ) =


0, even n _= 0,
−1
inπ
, odd n,
c0 =
1
2l
_ 2l
l
dx =
1
2.
(8.4)
Section 8 Other Intervals 363
Then,
f(x) =
1
2
−1
iπ
(eiπx/l − e
−iπx/l +
1
3e3iπx/l − 1
3e
(8.5) −3iπx/l + ・ ・ ・ )
=
1
2
−2
π
_sin πx
l
+
1
3
sin
3πx
l
+ ・ ・ ・_
.
PROBLEMS, SECTION 8
1 to 9. In Problems 5.1 to 5.9, define each function by the formulas given but on the
interval (−l, l). [That is, replace }π by }l and }π/2 by }l/2.] Expand each function in
a sine-cosine Fourier series and in a complex exponential Fourier series.
10. (a) Sketch several periods of the function f(x) of period 2π which is equal to x
on −π < x < π. Expand f(x) in a sine-cosine Fourier series and in a complex
exponential Fourier series.
Answer: f(x) = 2(sinx − 1
2 sin 2x + 1
3 sin 3x − 1
4 sin 4x + ・ ・ ・ ).
(b) Sketch several periods of the function f(x) of period 2π which is equal to x
on 0 < x < 2π. Expand f(x) in a sine-cosine Fourier series and in a complex
exponential Fourier series. Note that this is not the same function or the same
series as (a).
Answer: f(x) = π − 2
∞X
1
sin nx
n
.
In Problems 11 to 14, parts (a) and (b), you are given in each case one period of a function.
Sketch several periods of the function and expand it in a sine-cosine Fourier series, and in
a complex exponential Fourier series.
11. (a) f(x) = x2, −π < x < π; (b) f(x) = x2, 0 < x < 2π.
12. (a) f(x) = ex, −π < x< π; (b) f(x) = ex, 0 < x < 2π.
13. (a) f(x) = 2 − x, −2 < x< 2; (b) f(x) = 2 − x, 0 < x < 4.
14. (a) f(x) = sinπx, −1
2<x<1
2; (b) f(x) = sinπx, 0 < x< 1.
15. Sketch (or computer plot) each of the following functions on the interval (−1, 1) and
expand it in a complex exponential series and in a sine-cosine series.
(a) f(x) = x, −1 < x < 1.
Answer: f(x) =
2
π
∞X
1
(−1)n+1 sin nπx
n
.
(b) f(x) =
(
1 + 2x, −1 < x< 0,
1 − 2x, 0 < x< 1.
Answer: f(x) =
8
π2
∞X
odd n=1
cos nπx
n2 .
(c) f(x) =
(
x + x2, −1 < x < 0,
x − x2, 0 < x < 1.
Answer: f(x) =
8
π3
∞X
odd n=1
sin nπx
n3
364 Fourier Series and Transforms Chapter 7
Each of the following functions is given over one period. Sketch several periods of the
corresponding periodic function and expand it in an appropriate Fourier series.
16. f(x) = x, 0< x < 2. Answer: f(x) = 1 − 2
π
∞X
1
sin nπx
n
.
17. f(x) =
(
0, −1 < x < 0,
1, 0 < x < 3.
18. f(x) = x2, 0 < x < 10.
19. f(x) =
(
0, −1
2 < x < 0,
x, 0 < x < 1
2.
20. f(x) =
(
x/2, 0 < x < 2,
1, 2 < x < 3.
21. Write out the details of the derivation of the formulas (8.3).
9. EVEN AND ODD FUNCTIONS
An even function is one like x2 or cos x (Figure 9.1) whose graph for negative x is
just a reflection in the y axis of its graph for positive x. In formulas, the value of
f(x) is the same for a given x and its negative; that is
(9.1) f(x) isevenif f(−x) = f(x).
Figure 9.1
Figure 9.2
An odd function is one like x or sin x (Figure 9.2) for which the values of f(x)
and f(−x) are negatives of each other. By definition
(9.2) f(x) isoddif f(−x) = −f(x).
Notice that even powers of x are even, and odd powers of x are odd; in fact, this
Section 9 Even and Odd Functions 365
is the reason for the names. You should verify (Problem 14) the following rules for
the product of two functions: An even function times an even function, or an odd
function times an odd function, gives an even function; an odd function times an
even function gives an odd function. Some functions are even, some are odd, and
some (for example, ex) are neither. However, any function can be written as the
sum of an even function and an odd function, like this:
f(x) =
1
2
[f(x) + f(−x)] +
1
2
[f(x) − f(−x)];
the first part is even and the second part is odd. For example,
ex =
1
2
(ex + e
−x) +
1
2
(ex − e
−x) = coshx + sinhx;
cosh x is even and sinh x is odd (look at the graphs).
Integrals of even functions or of odd functions, over symmetric intervals like
(−π, π) or (−l, l), can be simplified. Look at the graph of sin x and think about
_π
−π sinxdx. The negative area from −π to 0 cancels the positive area from 0 to π,
so the integral is zero. This integral is still zero for any interval (−l, l) which is
symmetric about the origin, as you can see from the graph. The same is true for
any odd f(x); the areas to the left and to the right cancel. Next look at the cosine
graph and the integral _ π/2
−π/2 cosxdx. You see that the area from −π/2 to 0 is the
same as the area from 0 to π/2. We could then just as well find the integral from
0 to π/2 and multiply it by 2. In general, if f(x) is even, the integral of f(x) from
−l to l is twice the integral from 0 to l. Then we have
(9.3) _ l
−l
f(x) dx =


0 if f(x) is odd,
2_l
0
f(x) dx if f(x) is even.
Suppose now that we are given a function on the interval (0, l). If we want to
represent it by a Fourier series of period 2l, we must have f(x) defined on (−l, 0)
too. There are several things we could do. We could define it to be zero (or, indeed,
anything else) on (−l, 0) and go ahead as we have done previously to find either an
exponential or a sine-cosine series of period 2l. However, it often happens in practice
that we need (for physical reasons—see Chapter 13) to have an even function
(or, in a different problem, an odd function). We first sketch the given function on
(0, l) (heavy lines in Figures 9.3 and 9.4). Then we extend the function on (−l, 0)
to be even or to be odd as required. To sketch more periods, just repeat the (−l, l)
sketch. (If the graph is complicated, it is helpful to trace it with a finger of one
Figure 9.3
366 Fourier Series and Transforms Chapter 7
hand while you use the other hand to copy exactly what you are tracing. Turn the
paper upside down to avoid crossing hands.)
Figure 9.4
For even or odd functions, the coefficient formulas for an and bn simplify. First
suppose f(x) is odd. Since sines are odd and cosines are even, f(x) sin(nπx/l) is
even and f(x) cos(nπx/l) is odd. Then an is the integral, over a symmetric interval
(−l, l), of an odd function, namely f(x) cos(nπx/l); an is therefore zero. But bn is
the integral of an even function over a symmetric interval and is therefore twice the
0 to l integral. We have:
(9.4) If f(x) is odd,


bn =
2
l
_l
0
f(x) sin nπx
l
dx,
an = 0.
We say that we have expanded f(x) in a sine series (an = 0 so there are no cosine
terms). Similarly, if f(x) is even, all the bn’s are zero, and the an’s are integrals of
even functions. We have:
(9.5) If f(x) is even,
 
an =
2
l
_l
0
f(x) cos nπx
l
dx,
bn = 0.
We say that f(x) is expanded in a cosine series. (Remember that the constant term
is a0/2.)
You have now learned to find several different kinds of Fourier series that represent
a given function f(x) on, let us say, the interval (0, 1). How do you know
which to use in a given problem? You have to decide this from the physical problem
when you are using Fourier series. There are two things to check: (1) the basic period
involved in the physical problem; the functions in your series should have this
period; and (2) the physical problem may require either an even function or an odd
function for its solution; in these cases you must find the appropriate series. Now
consider f(x) defined on (0, 1). We could find for it a sine-cosine or an exponential
series of period 1 (that is, l = 1
2 ):
f(x) =
∞
_
−∞
cne2inπx where cn = _ 1
0
f(x)e
−2inπx dx.
Section 9 Even and Odd Functions 367
(The choice between sine-cosine and exponential series is just one of convenience in
evaluating the coefficients—the series are really identical.) But we could also find
two other Fourier series representing the same f(x) on (0, 1). These series would
have period 2 (that is, l = 1). One would be a cosine series
f(x) =
∞
_
n=0
an cos nπx, an = 2_ 1
0
f(x) cosnπx dx, bn = 0,
and represent an even function; the other would be a sine series and represent an
odd function. In the problems, you may just be told to expand a function in a
cosine series, say. You must then see for yourself what the period is when you have
sketched an even function, and so choose the proper l in cos(nπx/l) and in the
formula for an.
Example. Represent f(x) =
1, 0 < x < 1
2,
0, 1
2 < x < 1,
by (a) a Fourier sine series, (b) a Fourier cosine series, (c) a Fourier series (the last
ordinarily means a sine-cosine or exponential series whose period is the interval over
which the function is given; in this case the period is 1).
Figure 9.5
(a) Sketch the given function between 0 and 1. Extend it to the interval (−1, 0)
making it odd. The period is now 2, that is, l = 1. Continue the function with
period 2 (Figure 9.5). Since we now have an odd function, an = 0 and
bn =
2
1
_1
0
f(x) sinnπx dx = 2_ 1/2
0
sin nπx dx
=−2
nπ
cos nπx

1/2
0
=−2
nπ
_cos nπ
2
− 1_ ,
b1 =
2
π
, b2 =
4
2π
, b3 =
2
3π
, b4 = 0, ・ ・ ・ .
Thus we obtain the Fourier sine series for f(x):
f(x) =
2
π
_sin πx +
2 sin2πx
2
+
sin 3πx
3
+
sin 5πx
5
+
2 sin6πx
6
+ ・ ・ ・_
.
368 Fourier Series and Transforms Chapter 7
Figure 9.6
(b) Sketch an even function of period 2 (Figure 9.6).
Here l = 1, bn = 0, and
a0 = 2_ 1
0
f(x) dx = 2_ 1/2
0
dx = 1,
an = 2_ 1
0
f(x) cos nπx dx =
2
nπ
sin nπx

1/2
0
=
2
nπ
sin nπ
2.
Then the Fourier cosine series for f(x) is
f(x) =
1
2
+
2
π
_cos πx
1
− cos 3πx
3
+
cos 5πx
5
− ・ ・ ・_
.
Figure 9.7
(c) Sketch the given function on (0, 1) and continue it with period 1 (Figure 9.7).
Here 2l = 1, and we find cn as we did in the example of Section 8. As in that
example, the exponential series here can then be put in sine-cosine form.
cn = _ 1
0
f(x)e
−2inπx dx = _ 1/2
0
e
−2inπx dx
=
1 − e−inπ
2inπ
=
1 − (−1)n
2inπ
=


1
inπ
, nodd,
0, neven _= 0.
c0 = _ 1/2
0
dx =
1
2.
f(x) =
1
2
+
1
iπ
(e2iπx − e
−2iπx +
1
3e6πix − 1
3e
−6πix + ・ ・ ・ )
=
1
2
+
2
π
_sin 2πx +
sin 6πx
3
+ ・ ・ ・_
.
Section 9 Even and Odd Functions 369
Alternatively we can find both an and bn directly.
a0 = 2_ 1
0
f(x) dx = 2_ 1/2
0
dx = 1.
an = 2_ 1/2
0
cos 2nπx dx = 0.
bn = 2_ 1/2
0
sin 2nπx dx =
1
nπ
(1 − cos nπ) =
1
nπ
[1 − (−1)n].
b1 =
2
π
, b2 = 0, b3 =
2
3π
, b4 = 0, ・ ・ ・ .
There is one other very useful point to notice about even and odd functions. If
you are given a function on (−l, l) to expand in a sine-cosine series (of period 2l)
and happen to notice that it is an even function, you should realize that the bn’s
are all going to be zero and you do not have to work them out. Also the an’s can
be written as twice an integral from 0 to l just as in (9.5). Similarly, if the given
function is odd, you can use (9.4). Recognizing this may save you a good deal of
algebra.
Differentiating Fourier Series Now that we have a supply of Fourier series for
reference, let’s discuss the question of differentiating a Fourier series term by term.
First consider a Fourier series in which an and bn are proportional to 1/n. Since
the derivative of 1
n sin nx is cos nx (and a similar result for the cosine terms), we see
that the differentiated series has no 1/n factors to make it converge. Now you might
suspect (correctly) that if you can’t differentiate the Fourier series, then the function
f(x) which it represents can’t be differentiated either, at least not at all points.
Turn back to examples and problems for which the Fourier series have coefficients
proportional to 1/n and look at the graphs (or sketch them). Note in every case
that f(x) is discontinuous (that is, has jumps) at some points, and so can’t be
differentiated there. Next consider Fourier series with an and bn proportional to
1/n2. Ifwedifferentiate such a series once, there are still 1/n factors left but we can’t
differentiate it twice. In that case we would (correctly) expect the function to be
continuous with a discontinuous first derivative. (Look for examples.) Continuing,
if an and bn are proportional to 1/n3, we can find two derivatives, but the second
derivative is discontinuous, and so on for Fourier coefficients proportional to higher
powers of 1/n. (See Problems 26 and 27.)
It is interesting to plot (by computer) a given function together with enough
terms of its Fourier series to give a reasonable fit. In Section 5 we did this for
discontinuous functions and it took many terms of the series. You will find (see
Problems 26 and 27) that the more continuous derivatives a function has, the fewer
terms of its Fourier series are required to approximate it. We can understand this:
The higher order terms oscillate more rapidly (compare sin x, sin2x, sin10x), and
this rapid oscillation is what is needed to fit a curve which is changing rapidly
(for example, a jump). But if f(x) has several continuous derivatives, then it is
quite “smooth” and doesn’t require so much of the rapid oscillation of the higher
order terms. This is reflected in the dependence of the Fourier coefficients on a
power of 1/n.
370 Fourier Series and Transforms Chapter 7
PROBLEMS, SECTION 9
The functions in Problems 1 to 3 are neither even nor odd. Write each of them as the sum
of an even function and an odd function.
1. (a) einx (b) xex
2. (a) ln |1 − x| (b) (1 + x)(sin x + cos x)
3. (a) x5 − x4 + x3 − 1 (b) 1 + ex
4. Using what you know about even and odd functions, prove the first part of (5.2).
Each of the functions in Problems 5 to 12 is given over one period. For each function,
sketch several periods and decide whether it is even or odd. Then use (9.4) or (9.5) to
expand it in an appropriate Fourier series.
5. f(x) =
(
−1, −π < x < 0,
1, 0 < x < π.
6. f(x) =
(
−1, −l < x< 0,
1, 0 < x < l.
Answer: f(x) =
4
π
„
sin
πx
l
+
1
3
sin
3πx
l
+
1
5
sin
5πx
l
+・・・
«
.
7. f(x) =
(
1, −1 < x < 1,
0, −2 < x < −1 and 1< x < 2.
8. f(x) = x, −π
2
<x<
π
2
.
9. f(x) = x2, −1
2<x<1
2.
Answer: f(x) =
1
12
−1
π2
„
cos 2πx − 1
22 cos 4πx +
1
32 cos 6πx − ・ ・ ・
«
.
10. f(x) = |x|, −π
2
<x<
π
2
.
11. f(x) = coshx, −π < x < π.
Answer: f(x) =
2 sinhπ
π
„
1
2
−1
2
cos x +
1
5
cos 2x − 1
10
cos 3x +
1
17
cos 4x − ・ ・ ・
«
.
12. f(x) =
(
x +1, −1 < x < 0,
x − 1, 0 < x < 1.
13. Give algebraic proofs of (9.3). Hint: Write
Rl
−l =
R0
−l +
Rl
0, make the change of
variable x = −t in
R0
−l, and use the definition of even or odd function.
14. Give algebraic proofs that for even and odd functions:
(a) even times even = even; odd times odd = even; even times odd = odd;
(b) the derivative of an even function is odd; the derivative of an odd function is
even.
15. Given f(x) = x for 0 < x < 1, sketch the even function fc of period 2 and the odd
function fs of period 2, each of which equals f(x) on 0 < x < 1. Expand fc in a
cosine series and fs in a sine series.
Answer: fc(x) =
1
2
−4
π2
„
cos πx +
1
32 cos 3πx + ・ ・ ・
«
,
fs(x) =
2
π
„
sin πx − 1
2
sin 2πx +
1
3
sin 3πx − ・・ ・
«
.